How Long to Wait Before Attributing a Subscription to Onboarding

When you run an onboarding experiment, you're testing whether changes to the first-time user experience drive more subscriptions. The attribution window defines how long after that first session you'll count a purchase as influenced by the onboarding variant. Too short, and you miss conversions from users who need time to evaluate. Too long, and you attribute purchases that happened for completely unrelated reasons.

The window you choose directly affects which variant appears to win. A variant that creates immediate conversions might lose to one that merely postpones the purchase decision if your window is too generous. This is particularly problematic when you're testing aggressive tactics like immediate paywalls versus gradual value demonstration.

Seven days captures the overwhelming majority of onboarding-influenced conversions across most app categories. Users typically decide whether an app delivers value within their first few sessions, which usually occur in the first week. If they haven't subscribed by day seven, they're often waiting for a specific trigger: a reminder notification, a feature unlock, or simply forgetting about the app until a push brings them back.

Data from subscription apps shows conversion rates drop sharply after the first three days, then level off. The conversions that trickle in during weeks two through four rarely correlate with onboarding variant—they're responding to other variables like email campaigns, usage milestones, or seasonal factors. Attributing these to onboarding creates false confidence in variants that didn't actually drive the behavior.

Apps with long evaluation cycles may justify 14 or even 30-day windows. Fitness apps where users need to complete a workout program, budgeting tools that require a full month of data, or meditation apps building a daily habit might see legitimate onboarding-influenced conversions in week two or three. The key question is whether users are still actively engaging with the features introduced during onboarding.

Even in these cases, segment your attribution analysis by conversion day. If 80% of conversions happen in days zero through seven and the distribution doesn't differ meaningfully by variant, the longer tail isn't providing useful signal. You're just waiting longer to make the same decision you could have made with a shorter window.

It's tempting to extend attribution windows when early results don't reach statistical significance. More conversions mean more data, which could help a winner emerge. But this approach confuses two different questions: did the onboarding variant work, and did users eventually subscribe? These aren't the same thing, and conflating them leads to shipping variants that didn't actually improve the metric you care about.

Longer windows also slow down your experiment velocity. If you wait 30 days to attribute each cohort's conversions, you need at least 30 days before reading results, plus additional time to reach significance. That's 30+ days before you can start your next test. Shorter windows let you iterate faster, testing more hypotheses in the same calendar time.

Configure your attribution window before launching the experiment, not after seeing results. Changing the window based on preliminary data is a form of p-hacking that invalidates your statistical tests. Document the decision and the reasoning so future team members understand why that specific window was chosen for this specific experiment.

Tools like Spinosis handle attribution automatically when RevenueCat webhook events arrive with App User IDs that match experiment user IDs. Set your attribution window to match your hypothesis about onboarding impact, then trust the data that arrives within that window. If you're genuinely unsure, run a preliminary analysis of historical conversion timing before designing the experiment.